Gengo is a people-powered translation platform which allows customers to order via their website of API. Gengo automatically allocates orders among thousands of qualified translators around the world. Translators work online using Gengo’s simple and powerful translation tools.
As Product Manager I was responsible for maintaining and executing the product roadmap/strategy working closely with engineering, operations, sales and marketing teams in Tokyo, San Francisco and the Philippines. My role also encompassed user experience and quality assurance.
- Responsible for the specification, design, planning, testing and launch of:
- New responsive translation workbench with autosave and error detection
- Crowdsourced translation memory service with visual match differentials
- Consolidated commenting system with photo/video/audio support
- Integrated credit card payments, leading to 20% conversion rate increase
- Modular backend management web application for entire service
- Instigated new Style Guide to improve brand consistency and development efficiency
- Restructured scrum process, built offshore QA team and dramatically improved stability
- Decommissioned underperforming services and refactored areas of technical debt
The workbench is the heart of Gengo where translators do their work. Our goal was to provide a platform for translation that gives assistance when wanted, but otherwise gets out of the way and lets translators do what they do best.
Over a nine month period I looked at the way translators interacted with existing tools and worked with our design and front end teams to build a number of prototypes which were tested with real translators using different language pairs in order to validate the various hypothesis. Our findings formed the basis of the specification which I then broke down for implementation.
After launching the initial version I worked with our community manager to collect user feedback, in addition to quality and speed metrics, which was used to further iterate and improve the experience.
When translating things like product descriptions, accuracy is key in ensuring the translation faithfully represents what’s being sold. Mistranslating a dress as being size 2 instead of 12 could spell disaster and customer dissatisfaction.
In order to offset some of the risks that this can pose we designed and built a number of validation checks which operate in the background to find and highlight potential issues between the source and target text while the translator works (similar to spell check).
Depending on the certainty and severity of the issue, we can prevent submission of the job until the translator resolves the problem. This also allows us to preemptively catch individuals who may be trying to abuse the system (e.g. by using machine translation).
After its release, Gengo saw a drop in customer revision requests and using both quantitative analysis and qualitative feedback we made iterative improvements to reduce false positives and increase accuracy.
Translation memory (TM) is a database that stores previously translated content and makes it available for later reuse. This has the benefits of improving productivity and consistency while reducing costs for customers who don’t pay for reused translations.
While it had been used by the traditional industry for many years within offline tools, it had not been deployed at scale in a crowd-sourced environment. Similar to moving from a desktop word processor to an online document editor (e.g. Microsoft Word to Google Docs), at Gengo we set out to develop a system which would provide a real-time collaborative translation memory that is constantly updated and self-policing.
Its introduction has been a key driver for winning and retaining new business while leveraging an existing asset (the database of translations) and aiding knowledge sharing between translators.
To provide accurate translations, comprehension of the source content is key. In situations where words like “bike” could be interpreted as a “bicycle” or “motorbike”, more information is required to make a quick distinction.
In order to reduce ambiguity for context-sensitive content like product descriptions, user reviews and stock photography, we added the ability for customers to embed photos, videos and audio clips within comments.
We saw a rapid uptake from e-commerce customers of this feature with widespread praise by translators who no longer needed to ask for clarification when descriptions were unclear, ultimately improving the quality of the translation.
Credit Card Payments
Providing hassle-free and secure payment options plays a critical role in determining conversion rates when it comes time for a customer to place an order.
Having used PayPal since its inception, Gengo realised that it was losing customers due to the lack of direct credit card payments. In order to rectify this, without having to take on the burden of compliance requirements, we choose to integrate with Stripe who provide a simple and secure solution.
The introduction of credit card payments led to a 20% rise in conversion rates.
Managing translation at scale requires a complex system of both human and automated processes which were previously controlled by a number of disparate online and offline systems which were no longer fit for purpose.
I designed a new unified backend based on a system of easily extensible templates and modules which were implemented and released in discrete pieces. A flexible authentication system was put in place to provide fine-grained access permissions to employees, support agents and various contractors. Numerous time and cost savings were achieved as a result.
In an effort to help us build consistently and quickly without needing oversight for every pixel, I worked with our design and front-end teams, to develop a style or pattern guide. This helped us focus more on workflows and logic rather than individual components like buttons and list items.
The style guide was built on a responsive framework which allows the product to fluidly adapt to the myriad of different devices which our customers and translators use and the environment in which they live.
Since launch, it’s become a living reference tool for developers, designers, marketers, and others.